Sign in with
Sign up | Sign in
Your question

There's a new Sheriff in town; Real Temp 2.24

Last response: in Overclocking
Share
a b à CPUs
a b K Overclocking
March 20, 2008 6:11:13 PM

Most overclocking enthusiasts clearly understand that processor temperatures are critical, but which temps? CPU or Cores or both? What's the difference? How do they relate to one another? How do you know if they're accurate?

Many users may not realize that popular temperature monitoring utilities such as Core Temp, Everest and Hardware Monitor over-estimate Core temperatures (Tjunction), because they all "guess" at the unknown value of "Tjunction Max". Users may also not be aware that Intel does not support or publicly document Maximum Junction Temperature (Tjunction Max) on Core 2 desktop processors.

To deal with this problem more effectively, there's a new Sheriff in town; Real Temp 2.24 http://www.xtremesystems.org/forums/showthread.php?t=17...

Please read at least the first post, which has the link to download the program. It also explains how Real Temp works, and the temperature monitoring problems it solves. The thread has exploded to 20 pages. It's a bit of a read, but lends fresh perspective and understanding to the perpetual Tjunction Max / Core temperature debate.

Real Temp contends that all the popular temperature utilities such as Core Temp report excessively high values on many processors, so depending upon which variant you're running, Real Temp will typically report Cores at least 5c cooler than other utilities. I strongly agree with Real Temp, however, I think it could be perfected if additional variables were considered.

CPU temperature (Tcase) is the only temperature Intel specifies and supports. Although Everest and Hardware Monitor also report CPU temperature, Core Temp of course does not. SpeedFan reports both CPU and Core temperatures, but SpeedFan http://www.almico.com/speedfan.php is the only popular temperature monitoring utility which can be "calibrated".

This is precisely why I feature SpeedFan in the Core 2 Quad and Duo Temperature Guide http://www.tomshardware.com/forum/221745-29-core-quad-t... as the preferred temperature monitoring utility. The calibration methods are now based upon the use of known values for CPU offsets as well as Core offsets. If you follow the Guide to the letter, you'll have accurate temperatures in SpeedFan, which will report CPU temp slightly higher, and Core temps slightly lower than popular utilities.

Regardless, I see Real Temp as a very significant breakthrough in the ongoing effort to develop a simple temperature monitoring utility which may be more accurate from Idle through Load. I recommend checking out Real Temp.

Comp :sol: 

More about : sheriff town real temp

March 20, 2008 6:30:09 PM

comp,

So you saying the Core Temp program I am running generally over estimates the temp. then under estimates? BTW.. you posting is very informational.
a b à CPUs
a b K Overclocking
March 20, 2008 6:36:09 PM

Yes, that's correct. It means that Core temperatures are actually a bit cooler than you've all been led to believe. Just download Real Temp and check it out for yourself.
Related resources
March 20, 2008 6:45:27 PM

ok so speedfan if calibrated give the best results..
Core Temp is too high..

Where do you think Real Temp falls? I am guessing it is more accurate then Core Temp but doesn't to tend to favor one side.

Like over or under estimate?
a b à CPUs
a b K Overclocking
March 20, 2008 6:50:19 PM

Core Temp = higher
SpeedFan = intermediate (only if calibrated)
Real Temp = lower
March 20, 2008 6:53:24 PM

Interesting your user's configuration was a good read and well done.. ;-) So what were your temps. when running @ 3.8 with 1.536 Vcore. Not sure if that was voltage in windows or in the bios.
a b à CPUs
a b K Overclocking
March 20, 2008 7:00:10 PM

Prime95 Small FFT's at 3.7 indicates 68c CPU and 75c average Core, which max's out my rig, so I don't run load tests at 3.8. Vcore is in Windows during Load. My 24/7 settings are in my sig, which are 3.6, 62c CPU, 68 average Core, 1.440 Vcore.
March 20, 2008 7:18:50 PM

What was the VID for you Q6600 GO?. I happened to get one with VID of 1.325. So in order to get my Q6600 up to 3.6 I had to run the vCore above 1.5 to be stable in windows Vista 64. To keep it cool I went with Xigmatek HDT-S1283 and Kaze fan, that is pretty loud.


http://www.tomshardware.com/forum/page-249073_31_120.ht...
a b à CPUs
a b K Overclocking
March 20, 2008 7:24:59 PM

1.3000
March 20, 2008 9:59:29 PM

Very interesting. After getting a new pc, I didn't want to set up speedfan so I just grab Coretemp.

Anyways, since Coretemps reads higher temps, it's pretty much safer for your system. I think I rather continue to be fooled by higher temps from Coretemps. :) 

Of course I'm gonna grab this and see how much lower the temps really are.
March 21, 2008 2:00:59 AM

I can't get the program to work. says the zip file is invalid. Anybody else get it to work?
March 21, 2008 3:40:36 AM

It worked off the top over here, just dunno how to set it up!

--Lupi
a b à CPUs
a b K Overclocking
March 21, 2008 3:47:16 AM

Initially, it doesn't need any setup. If you read the first post in the thread, which will eventually be included in the readme file, you'll see that one of the features is that it allows you to calibrate your Idle temps if they happen to be below ambient, while not disturbing the Load temps. Since some processors are known to have non-linear Idle temps, this can be a very helpful.
March 21, 2008 3:54:01 AM

I tried and tried, the developer of RealTemp thinks almost every CPU has a 85c Tj max even quad cores. He thinks adding a idle offset magicaly makes your temps not below ambient. Changing the numbers displayed on a program don't change the cores' temp, he also says Tcase is not important and should be "ignored".... while Intel clearly rates thermal spec on Tcase and not core temps.

His findings with the IR gun and the graph about the "proposed" inaccurate DTS is interesting although wrong.. trying to explain it to him is like talking to a wall. He kept insisting the temp he gets from the IR gun reading is fail proof and accurate temps of the DTS, he doesn't understand that hes only getting IHS temp which is... yes Tcase or pretty close to it. Even if the IHS was removed and the IR gun placed on the die it stil won't be accurate as the heat is orignating right within the core, a spot an IR gun cannot get to. The heat has to spread throughout the die and into the IHS, thus the temp hes getting will be lower most likely by about 5c which is where the delta comes in.

Reguarding the DTS graph I also fail to believe any of that especially with the 65nm CPUs, any common sence will tell you Intel will not mass produce millions of processors with defective DTS such as he described (this is directed to 65nm's), obviously they would fix it. Although many 45nm's have a "stuck" DTS the method of idle offsets in RealTemp does NOT help "compensate" for the errorous DTS in anyway. Any processor with a sticking DTS should be RMA'ed anyway, but changing a displayed number to make it higher or lower (usually higher) just to make it "look sensable" is totally wrong.

Thats equal to saying the temp of my oven is 300c, but iam going to use an offset to make it display at 315c. Yea......

Right now I highly reccomend Speedfan so you can correctly set the Tj max for your processor, CoreTemp is also inaccurate but is much much closer to being accurate, I already posted a large post explaining why each stepping is what Tj max and the developer said he "will consider it" so.. I hope he does but if he don't I or we always got Speedfan thats totally customizable.

RealTemp was almost perfectly accurate in its early 1st and 2nd release versions, but now its just all inaccurate. It still has potiential if the Tj max values get fixed up thats all it needs, other than that I like it. I personally use version 2.11 as its the last accurate one for most CPUs. As always some agree to it and some don't.. if you really believe it then so be it, I can't change your mind for you :) 
a b à CPUs
a b K Overclocking
March 21, 2008 4:04:32 AM

gigabyte, I agree with you. It's not possible to measure junction temperatures within the nano-architcture of the DTS arrays with an IR device. Whatever the estimated Core temperatures are, according to whichever temperature monitoring utility, there is a know relationship between the CPU temp and the Core temps, which remains constant. We can use this information to get accurate temps. Here's how I do it in my Temp Guide.

We know that CPU temperature has to be higher than ambient, which we can measure. We also know that Idle power dissipation and cooler efficiency both have documented values. If we duplicate stock conditions for Idle power dissipation, then we can calibrate the CPU temp more accurately than blindly trusting the canned values coded into BIOS. That sounds great, but we can't calibrate any temps unless we have a program which allows offset corrections, which is SpeedFan.

As Evilonigiri pointed out, SpeedFan takes a little time to set up, but it's not rocket science, and since SpeedFan can be calibrated, I think it's well worth the effort. OK, so if we calibrate the CPU temp in SpeedFan, how do we know what the Core temps should actually be? I've read and studied hundreds of pages of Intel papers, and I've tested dozens of processor vaiants, some of which makes no sense. The document which I consider to be the thermal "Holy Grail" is the following: http://eda-publishing.imag.fr/spip/IMG/pdf/TMI23.pdf

As I know you're aware, this appears to be the only Intel document which reveals the unknown factor, and strikes precisely at the root of what all the Core temp confusion is about. It discusses the relationships between Tcase (CPU temp) and Tjunction (Core temp), which shows that Core temp peaks at 5c higher than CPU temp under Load. With this factor in mind, if we again duplicate stock conditions, then we can use calibrated CPU temp, which in turn can be used to calibrate Core temps.

This means that Tjunction must include the following variables; measured ambient, Idle power dissipation, (which is series and stepping dependent), cooler efficiency, and Tcase to Tjunction Delta 5c at Load. So to simplify, Core temps are determined by Ambient to Tcase to Tjunction Deltas, which means that Tjunction Max is actually a "floating" value. This may explain why Intel does not document or support Tjunction Max for determining Core temps.

Comp :sol: 
March 21, 2008 1:06:15 PM

I'll stick with everest and the odd running of coretemp. There is no real need to fudge another 5c down just for the sake of thermal limitations especially with a lot of users wanting to use the 45nm chips where thermal limits are well within tolerance. The future problems (fried chips) will be due to overvolting not temps.

I guess the q6600 (which I have) overclocks will still be thermal limited, but as people upgrade from it to an affordable 45nm quad with a 9+ multi heat won't be the issue; electron migration will. At any rate I'm finding myself less concerned with heat as air will be the norm for overclocking 45nm chips with water being less and less needed.

I will admit the sub-ambients (LN, DI, PE) still gives you the benefit of improved internal chip performance and reduced electron migration (ie, higher volts, higher clocks, less damage).
March 21, 2008 8:01:38 PM

Core temp reads my two cores on my x2 4000+ Brisbane at about 10c and 15c when idle. Are you telling me that Real Temp will read lower than that? I find it hard to believe that that would be accurate. I sometimes wonder how accurate my temps are, because I barely believe that my cores can run as low as they do.

But, I am always interested in the most accurate software possible.
a b à CPUs
a b K Overclocking
March 21, 2008 8:10:40 PM

If you click on the link to Real Temp in my first post, it says that Real Temp is designed for all Intel Core processors. However, one of the features of Real Temp is Idle temp calibration, which allows Cores with non-linear sensors known to incorrectly Idle below ambient, to be corrected without disturbing Load temps.
March 23, 2008 2:19:59 PM

Speed Fan CPU Idle : 25C
Real Temp CPU Idle : 29C
Core Temp CPU Idle : 39C

That's one hell of a difference... I'll stick to SpeedFan it makes me happier :p 
a b à CPUs
a b K Overclocking
March 23, 2008 6:35:51 PM

ThrillerGTX, Core Temp and Real Temp do not have a "CPU" temp. Only SpeedFan has a CPU Temp. You have obviously substituted the term "CPU" for the term "Core".

Also, in order for your "Core" temperatures to be 25c in SpeedFan, your Ambient temperature would have to be about 17c. Is your computer in a cold environment? Are you running SpeedFan 4.33 or SpeedFan Beta 4.34? Regardless, I doubt that your temperatures are correct. You can, however, calibrate your temperatures in SpeedFan by following my Temp Guide. There a link to it near the bottom of my first post. You might want to give it a try.

Comp :sol: 
!